The Sublimation of Metals at Low Pressures

58

The Siihlimation of Metals at Low Pressures.
By G. W. C. Kaye, B.A., D.Sc, and Donald Ewen, M.Sc.

(Communicated by Dr. E. T. Glazebrook, F.RS. Eeceived June 10, —

Bead June 26, 1913.)

(From the National Physical Laboratory.)

[Plate 5.]

Introductory.

Many metals have been found to exhibit evidences of volatility at
temperatures considerably below their melting points. As long ago as 1872,
Merget* demonstrated that frozen mercury volatilised perceptibly in air in
course of time. DemarQay,f in 1882, conducted similar experiments in vacuo
and found that cadmium evaporated sensibly at as low a temperature as 160^,
zinc at 184°, and lead and tin at 360° C. In 1887, Zenghelio| obtained
evidence of the volatility of lead, copper, zinc, etc., even at room temperatures.
Spring,§ in 1894, working at atmospheric pressure, showed that zinc was
appreciably volatile at 300°, and cadmium and copper at 500°. Eoberts- Austen
and Merrett, in some unpublished experiments at the Eoyal Mint, in 1896,
detected the volatility of cadmium and zinc at 100° in vacuo, while KrafftH in
1903 and 1905 investigated in some detail the volatilisation of a number of
metals at low pressures. Eosenhain, at the National Physical Laboratory,
has obtained beautiful crystals of sublimed zinc by heating a piece of zinc
to 300° C. for some weeks in a glass tube containing hydrogen at atmospheric
pressure.

The notable phenomena of the interdiffusion of metals, with which Eoberts-
Austen's name is associated, provide, of course, additional evidence of the
vapour pressure that solid metals exert even at ordinary temperatures.

A familiar illustration of metallic volatilisation is furnished by the

blackening of tungsten and carbonlF filament lamps. Deposits of definite

outline can occasionally be detected on the bulbs of the lamps ; a fact which

seems to point to the projection of particles in definite directions from the

filament.

* Merget, * Ann. Chim. Phys.,' 1872, vol. 25, p. 121.
t Demargay, * Comptes Rendus,' 1882, vol. 95, p. 183.
J Zenghelio, * Zeit. Phys. Chem.,' 1887, vol. 1, p. 219.
§ Spring, ^ Comptes Rendus,' 1894, p. 42.

II Krafft, ' Ber. Deut. Chem. Gesell.,' 1903, vol. 36, p. 1690, and 1905, vol. 38, p. 254.
IT Berthelot showed, in 1904, that such vaporised carbon is not graphitic, but amorphous.

The Suhlimation of Metals at Low Pressures.

59

The extent of the disintegration exhibited by a heated metal depends a
great deal on its nature. In the case of the metals of the platinum group,
Crookes* found that when they were heated in still air at atmospheric
pressure, they arranged themselves in the following order of increasing
volatility: Eh, Pt, Pd, Ir, and Eu. The marked volatility of iridium at
temperatures above 1000° C. has long been known to users of platinum-iridium
thermocouples.

The disintegration of a metal increases rapidly with a rise in the
temperature, and almost all observers are agreed that the presence of oxygen
serves to augment the effect, at any rate in those cases which have been
investigated. Hydrogen and nitrogen do not, in general, favour dis-
integration.

The platinum metals, with the exception of palladium, all disintegrate less
as the pressure is lowered, and accordingly it does not appear that in these
cases the effect is one of true sublimation. Eobertsf has recently conducted
experiments with these metals, and infers that the volatilisation is not a
simple process, but is brought about by the formation of endothermic oxides
more volatile than the metals themselves.^

With most other metals, however, we should naturally expect volatilisation
to be facilitated by a reduction of pressure. We have collected in the
following table, the data for a number of metals for which information

Metal.

Boiling point.

Volatilisation
detectable at

Melting point
at 1 atmos.

At 1 atmos.

In vacuo.

Mercury

Potassium

Sodium

Cadmium

Zinc

Bismuth

Lead

Silver

Copper

"0.
357
760
880
778
918
1420
1525
1955
2310
2270
2530 ?
2450
2500?

°C.
160
370
420
450
550

1000

1150

1400?

1600?

1700?

1800?

°C.

-39
63
97
160
180
269
360
680
400
360

1370
950

1200

°C.

-39

63

97

321

419

269

327

961

1084

232

1064

1500

1750

Tin

Gold

Iron

Platinum

■^ Crookes, * Eoy. Sec. Proc.,' May, 1912, A, vol. 86, p. 461.

t Eoberts, ' Phil. Mag.,' February, 1913, p. 270.

{ In this connection, see Goldstein (' Ber. Deut. Chem. Ges.,' 1904) and Magnus (* Phys.
Zeit.,' 1905, vol. 6, p. 12), who showed that Pt and Ir at a white heat rapidly absorb
oxygen.

60 Dr. G. W. C. Kaye and Mr. D. Ewen.

regarding the effect of pressure on the boiling point is available. The values
of the higher boiling points are largely due to Greenwood.^ Some of the
volatilisation temperatures quoted have been recently determined at the
National Physical Laboratory.

Columns 2 and 3 reveal the marked effect of pressure on the boiling point ;
and Column 4 shows the temperatures at which volatilisation has been
detected, mostly at low pressures. These temperatures are intended to imply
appreciable volatilisation ; if the experiments were sufficiently prolonged,
volatilisation could be detected in some cases at temperatures even lower
than those given, as will be gathered from p. 58. Column 5 gives the corre-
sponding melting points and is added for the sake of comparison.

Rectilmear Emission of Particles.

Evidence has recently been obtained of the emission of particles of metal
at right angles to the surface of a heated metal, in much the same way, for
example, as particles of metal are ejected from the surface of a cathode in a
discharge tube. In most cases, this straight line emission is obscured by
general volatilisation ; but when the circumstances were favourable its
existence has been detected.

Eeboul and de BoUemontf have recently shown that small strips of either
copper or silver, when heated in an electric furnace at temperatures from
400°^ to 900° C. yield black deposits which closely follow the outline of the
emitting metal. Thus when the latter was cut in the shape of a cross, the
deposit also was cruciform. The deposits were received on a platinum screen,
and proved to consist either of the emitting metal or its oxide. In air at
atmospheric pressure, 3 mm. was the greatest distance at which such deposits
were obtained ; the best results were obtained at about 1 mm. distance. In
oxygen, the effect was enhanced ; in a vacuum, the deposit gained in sharp-
ness of outline. Curiously enough, in hydrogen, the edges of the strip seemed
to be the only active regions, so that the sputtered image reproduced merely
the outlines of the strip.§

The rate of deposition increased very considerably as the furnace was made
hotter. On repeating the experiment with a sample of copper which had
already been used, the deposit was much less dense than that obtained on the

^ Greenwood, 'Eoy. Soc. Proc.,' 1909, A, vol. 82, p. 396 ; 1910, A, vol. 83, p. 483.

f Eeboul and de Bollemont, ^ Journ. de Phys.,' July, 1912, 5, vol. 2, p. 559.

J Below 400° no deposits were obtained.

§ A somewhat similar edge-effect was obtained by one of us in connection with the
sputtered deposit from an aluminium cathode in a discharge tube (Kaye, * Phys. Soc. Proc.,'
Feb., 1913, vol. 25, p. 198).

The Sublimation of Metals at Loiv Pressures, 61

first heating. Other metals — nickel, iron, and aluminium — were tried, but
without success.

Experimental.

The present authors, during the early part of last year, obtained somewhat
similar results which are recounted in this paper.

Iridium. — In one experiment, with which Dr. Harker was associated,* a
strip of pure iridium (S, fig. 1) was heated by the passage of a heavy alternat-
ing current of low voltage. The strip was arranged, edge upwards, within a

c

B

EiG. 2.

hollow metal cylinder of diameter about 18 mm. The gas was nitrogen and
the pressure about 20 mm. At the end of the experiment two horizontal
bands of deposit A and B were observed on the inner surface of the cylinder,
each one facing a side of the iridium strip, and of roughly the same width.
The rest of the cylinder was not wholly free from deposit, but it reached
a minimum at points Cand D opposite the edges of the strip.

Go'p'per, — In a second experiment, a copper tube (T, fig. 2) with a hole
(diameter 8 mm.) in its side was heated from within by a strip of metal
through which alternating heating current was passed. As shown in fig. 2,
the strip was not in contact with the cylinder. The pressure was about
1 mm. and the gas nitrogen.

After a few. minutes the copper attained a visibly red heat (800° C), and a
black deposit rapidly formed some 4 cm. away on an opal-glass plate (P)
placed obliquely, as indicated in the figure. The deposit approximated in
shape to an elliptical ring. This is to be ascribed to the sputtering action
of the edge of the hole in the tube. The outline of the sputtered ring
was shortly afterwards largely obscured by general deposit from the body
of the tube and the heating strip. Fig. 3 gives a notion of the appearance
of part of the elliptical band of deposit ; it has been strengthened slightly
in the photograph (see Plate 5).

* See Harker and Kaye, * Roy. Soc. Proc.,' 1913, A, vol. 88, p. 536.

62

Dr. G. W. 0. Kaye and Mr. I). Ewen.

Iron Strip A

IroQi. — In a further series of experiments/some interesting deposits have
been obtained with iron. Commercially pure Swedish wrought iron was
used in the form of thin strips which were highly polished. The arrange-
ment of the apparatus is indicated in fig. 4.

The iron strip A was mounted vertically and connected at its extremities
to two stout copper leads, by which means the strip was heated electrically

by direct current. Parallel to, and
at about 1 mm from, the polished
surface of the iron was stretched a
strip of platinum foil B, in the
centre of which was a small hole
about 3 mm. square. A second
piece of platinum foil C was welded
on to B, and contained another
similar hole so arranged that the
two holes in B and were opposite
each other and about 5 mm. apart.
A third strip of platinum foil D was
mounted opposite the hole in C and
about 1 mm. away from G. Strips
B and C were electrically insulated
from the iron strip A, and the
outermost strip D, which was
intended to receive the deposit, was
connected to the positive end of
the iron 'strip. The whole was placed in a vessel which was highly
exhausted, and the iron strip was then heated by passing through it a
direct current of about 20 to 30 amperes at 50 to 100 volts. The
temperature of the iron was kept at about 950° C. for some 5 hours.*

By this means, a dark brown deposit was obtained on the platinum strip D,
and, as shown in the photograph (fig. 5), the deposit took the form of a well
defined image of the square holes in the other pieces of foil. The clearly
defined edges of this shadow make it difficult to imagine that the method
of transference of the material can be other than some kind of rectilinear
propagation of the volatilised particles. The slight distortion of the image
is due in part to a want of alignment of the two holes in B and C, and
partly to the shape of the hole in G.

No deposit was found on the sides of B and G remote from the hot iron
strip. Fig. 5 shows the black deposit on the side of B facing the iron
•^ The temperature was measured by a hot-wire optical pyrometer.

3c tn.

Fig. 4.

The Sublimation of Metals at Low Pressures, 63

strip. There was also some deposit on the corresponding face of around
the sides of the hole ; a feature which indicates that not all the particles
are projected normally from the iron. In some experiments, plate C was
removed and only the one hole (in B) used ; the deposits in these cases
were not quite so well defined.

The surface of the emitting iron strip, when examined under a fairly high
power, shows regularly oriented etched pitting (fig. 6). The method of heating
by direct current seems to favour the development of these etched pits to an
extraordinary degree. As described later, specimens of iron were also heated
in a tube furnace in vacuo to the same temperature as before, viz., 950° C.
In these cases, the etched pitting of the surface, although generally
distinguishable, was not developed to anything like the same extent as in
specimens heated under the same conditions by the passage of the heating
current through the specimens themselves. It would appear, therefore,
that the electrical conditions which obtain in the latter case predispose
the metal to disintegration, as evidenced by the etching effects.* It would
be of interest to see if heating by the passage of an alternating current
through, a metal would also produce such marked pitting. The frequent
twinning, an example of which is shown in fig. 6, is, of course, characteristic
of the structure of iron at the temperature of the experiments.

The central and hottest portion of the iron strip, which was opposite the
holes in the platinum strips, was afterwards found to be brown in colour.
The appearance suggested that a certain amount of oxidation had taken place
in spite of the fairly high vacuum employed.f This is supported by the fact
that on subsequently annealing the iron in hydrogen the brown colour
disappeared and the iron reverted to its normal tint.

The deposit was found, when tested, to give a distinct iron reaction, thus
proving that the shadow was due to the transference of particles from the
iron. The loss in weight of the iron specimen was too small to be detected
by an ordinary chemical balance. The adhesion of the deposits to the receiv-
ing strip of platinum was remarkable, vigorous polishing for some minutes
being necessary for their removal. Under the microscope, the surface
of the platinum where the deposit was formed shows a reticular pattern
resembling the structure obtained on polished and etched metallic surfaces.

* Fredenhagen ('Phys. Zeit./ 1912, vol. 13, p. 539) found a parallel effect with the
negative electrical discharge from a hot metal in "vacuo. The emission from a certain
electrode, when the latter was heated in an electric furnace, was only 1 per cent, of that
obtained on heating by direct current to the same temperature.

t The pressure was initially of the order of 0*004 mm. of mercury and during the course
of the heating averaged about 0*035 mm., the rise being accounted for by the evolution of
gases from the heated iron.

64 Dr. G. W . 0. Kaye and Mr. D. Ewen.

As the result of a number of experiments, it was found that the maximum
range of the iron particles was about 1 cm. in a good vacuum ; at higher
pressures it would probably be less.*

Deposits from iron were also obtained on non-metallic receiving surfaces,
such as fused silica.

In order to examine more closely into the cause of this transference of
material, some further experiments were carried out with iron, in which the
arrangements were similar to those already described, but the heating was
effected by an electric tube-furnace, wound with a spiral resistor of nichrome
wire ; the iron specimen did not, therefore, in this case, carry any current.
The window in these experiments was a long slit parallel to and about
2 mm. from the strip ; only one window was employed. As before, quite a
sharp shadow of outline corresponding to the slit was obtained on a
platinum screen, about 1 mm. distant from the slit. This is shown in
fig. 7; the illumination of the screen in the photograph causes the grey
deposit to appear white on a dark ground.

With this arrangement, one hour's heating produced only very faint
indications of a deposit, whilst a clearly defined image was obtained from a
run lasting three hours. With direct-current heating, on the other hand,
the rate of deposition was much more rapid, and in one experiment,
10 minutes sufficed to produce a fairly clear deposit (see also p. 63).

In the tube-furnace experiment the temperature was not quite uniform,
so that both iron strip and platinum screen were a good deal hotter at one
end than the other. The deposit on the screen was found to be fainter at
the hotter end; this was probably due to the greater loss by evaporation
from the hot end of the screen. The deposit shown in fig. 7 did not possess
the brown colour of the shadows obtained in the previous experiments, nor,
as has already been remarked, did the surface of the iron show such charac-
teristic etched pitting. Thus, although more prominent deposits were
obtained when there was slight oxidation, it would appear that the presence
of oxygen — at any rate, in quantity sufficient to cause visible oxidation — is
not essential to the process of transference.

Tungsten. — Extensive deposits and corresponding shadow results were also
obtained when tungsten was heated in vacuo to about 1800° C.

Discussion,

From the foregoing results, and those obtained by Keboul and
de BoUemont, it would appear that there are two main classes of vapour
given out when a metal volatilises; one kind, which is associated with

* Cf. Eeboul and de BoUemont, above.

The Sublimation of Metals at Low Pressures. 65

evaporation as usually understood by the term, the other, made up of
particles of metal, which travel in straight lines from the surface of the
metal, which they leave approximately at right angles, and which, as our
experiments appear to show, have (in the case of iron, at any rate) a range
of only a centimetre or so m vacuo, "What may be the inherent cause of
difference between these two kinds of particles we are not in a position to
say at present ; we suggest that the " rectilinear '' type consists of electrified
particles of metal, while the ordinary vapour particles are electrically neutral.

It is well known that, if a liquid has its surface suddenly changed in area,
a surplus charge of electricity makes its appearance, as, for example, in the
splashing of water or mercury. On the same grounds, the pitting of the
heated surface, and the consequent alteration of area, would be expected to
release a certain amount of electrification, which would, in favourable cases,
accompany the liberated particles. The repulsion of the charged particles at
right angles to the surface of the heated metal would follow if the surface
were also charged with like sign ; this, of course, is possible with a strip
heated by direct current. We have already remarked on the special
efficacy of direct current in producing deposits.

The effect is not so easy to explain in the case of the tube-furnace and
alternating-current experiments. Eeboul and de Bollemont suggest, in
explanation of the phenomena, that the transference of the material is due
to miniature eruptions caused by the explosive combination of occluded
hydrogen and oxygen in the metal. On this view, the effect would be
expected to fatigue, and this is in accordance with their experiments (p. 60).
The explanation does not, however, seem to us convincing.

In our experiments with iron, the temperature did- not exceed about
1000° C, under which conditions we should expect that any electrified
particles there might be would carry a positive charge ; and, in fact, Sir J. J.
Thomson* showed some years ago that positively charged particles of metal
were among the positive ions given off by platinum heated to moderate
temperatures at low pressures. It would be interesting to see if a difference
could be detected in the intensity of the deposits obtained from either end of
a strip of metal heated by direct current. On the above hypothesis, the
positive end of the strip might give an appreciably heavier deposit at such
temperatures.

At higher temperatures than we have employed — near and above the
melting point of iron — negative electricity predominates, and opposite results
would accordingly be expected in such an experiment. The effect of a
magnetic field on the stream of particles would, of course, be a valuable

* * Conduction of Electricity througli G^se^/ 1906, p. 217.
VOL. LXXXIX. — A. F

66 Dr. G. W. C. Kaye and Mr. D. Ewen.

piece of evidence. By such means, Owen and Halsall*' find, however, that
in a good vacuum, and over a wide range of temperatures all high enough to
give negative ionisation, the thermionic current, in the case of Pt, Pd
and Ir, is due almost entirely to electrons. They conclude that the
proportion of heavy and metallic negative ions is certainly less than 1 part
in 2000. It should be remarked also that Eoberts, in the paper already
referred to, did not find any evidence of particles which were electrified in
the vapours of the platinum metals. But, from his published account, we
should gather that the fog-condensation chamber, by which he tested the
point, was probably too remote from the heated metal to detect such
short-range particles as we have described.

The influence of traces of oxygen in producing a kind of " weathering " of
the surface of the metal is one on which stress has been laid by a number of
experimenters, and it is reasonable to suppose that some such action would
augment the disintegration to a marked degree with some metals. It is
significant that most workers are agreed that the presence of oxygen
accentuates the positive electrical emission from hot metals ; in such cases
we may regard the charged particles as the direct outcome of the energy of
reaction between the metal and the gas. We have elsewhere noted Eoberts'
conclusions as to the part played by oxygen in the volatilisation of the
platinum metals, and it is a matter for further investigation to ascertain the
extent of the effect with the baser metals. Some evidence is afforded by the
experiments on p. 64.f

As an alternative to the charged-particle hypothesis it is not impossible
that the distinction between the " rectilinear " particles and the ordinary
particles is one chiefly of size. We should expect that very small emitted
particles of metal — with dimensions not far from molecular — would suffer
appreciable scattering by the gas molecules and lose very speedily their
original direction of projection. But larger particles, projected with the same
velocity, would travel farther before being similarly disturbed. Possibly the
range of the projected particles under the same temperature conditions varies
considerably from metal to metal ; and this may account for the lack of success
which attended Eeboul and de Bollemont's efforts to obtain deposits with
metals other than copper and silver at atmospheric pressure.

It may not be too far from the purpose of this paper to consider the
possible source of such large particles. Dr. Eosenhain and one of usij: have

^' <PML Mag.,' May, 1913, vol. 25, p. 735.

t See also Humfrey, ^ Iron and Steel Inst. Journ.,' 1912, Carnegie Memoirs.
J " Interciystalline Cohesion in Metals^" Eosenhain q.iad Ewen, ' Inst. Metals Journ.,
Bept., 1912,

Kaye and Etven.

Roy. Soc. Proc, A, vol. 89, Plate 5.

Fig. 3.- — Copper Deposit.

>ifim»0

B and C

D

Fig. 5.— Photograph of Hole and Iron

Deposit cast by it.
Jn the above figure, plate C is behind

plate B. Full size.

Fig. 7. — Iron Deposit. | full size.

Fig. 6.— Pitted Surface of Iron Strip, x 500 diameters.

The Sublimation of Metals at Low Pressures, 67

put forward a theory as to the mechanism by which evaporation takes place
from crystalline metals. The view was adopted, from the observations of the
behaviour of a number of metals in vacuo, that the volatilised metal consists
initially of the intercrystalline amorphous material which cements together
the crystal faces. This amorphous cement is more volatile than the crystals
themselves, and accordingly grooves or channels are formed along the crystal
boundaries of metals subjected to prolonged heating in vacuo. The increased
liability to erosion along the sides of these channels may perhaps be the cause,
directly or indirectly, of particles larger than those from the body of the
crystals. With polished specimens the channels are visible under the
microscope ; the process is known as vacuum etching.

If there is any real analogy between the sputtering from a cathode in a
low-pressure discharge tube, and thermal sputtering such as we have described,
it may be that cathodic sputtering carried out under suitable conditions
would similarly lead to the formation of patterns corresponding to the
structure of the metal.

So far as evaporation from the isrystals themselves is concerned, we may
usefully employ the conception of a '' crystal unit " adopted by some metallo-
graphists. The term implies a small ordered stable aggregate of molecules
which serves as a brick from which to build up the crystal structure. We
may imagine that either through the application of heat or by chemical
combination the stability of a surface unit is endangered by reason of the
loss of individual outlying molecules. The whole unit disintegrates and
comes away piecemeal from the crystal in particles, may be, of appreciable
size, and at the same time a pit is commenced in the crystal surface. We should
not be unreasonable in expecting that such particles coming from a stable
crystal system would be electrically charged, in contradistinction to those
coming from an unordered amorphous medium.

The results described in this paper are to be regarded as of a preliminary
character; it will be apparent that there is scope for further work in a
number of directions.

Fig. 3.- L'uppL-r Uupowit.

m

B ami C

Fk;. T). — Photogiapli of Hole ami Iron

Deposit wist by it

In tlie aliove iigiiic, plate C is luihiiid

piate K Fill! size.

Kui. (>.--I'itl<(i Siiif;ir(- of lion Snip, x 'i(X> diaiiinU'rs.

Fio. 7.— Iron Deposit. j-J full size.